ralph-wiggum

You can prevent context pollution by deliberately rotating to fresh memory contexts before accumulation builds up. This approach clears old context slices during long-running tasks and persists state externally in files and git.

Context rotation is a memory management technique that cycles to fresh context slices during multi-step workflows. It ensures state persistence in files and git, enabling traceability and survival across restarts.

Yes, you can use git to persist LLM state across iteration loops. The context rotation mechanism supports state persistence in git, allowing multi-step workflows to survive restarts while maintaining traceability.

You can recover from failures in autonomous iteration loops by using built-in guardrails. These guardrails guide progress and provide recovery mechanisms, while state persistence in files and git ensures no work is lost.

You should rotate memory context before pollution builds up in long-running AI tasks. This is especially applicable for complex, multi-turn workflows where accumulating irrelevant context degrades performance.